Thread: Driver or Postgres doesn't report refer. integr. errors
Hi,
I'm running 7.0 on NT (with cygwin) and ODBC 6.50
I picked up the pgupt.mdb (0.15) to upsize from Access 97.
I changed it to use ON UPDATE, ON DELETE ... in the Create statements, discarding all the other trigger creations. I'm still early in development, so I still have not many ideas on the interaction with the drivers.
The good news is that all my database with a lot of tables just convert correctly to 7.0 (including the whole referential integrity) and default values. Perhaps I will publish soon all the modifications I did on pgupt.mdb
After that, I linked all the tables in Access 97 (fast and easy).
The bad news, whenever I delete or update a record causing error in referential integrity, no error is reported to Access. The record seems to disappear, but when rereading the table (or recordset), the 'deleted' record reappears.
(please send cc to me: aerd@mx3.redestb.es )
Thanx in advance,
Erny
> The bad news, whenever I delete or update a record causing error in referential integrity, no error is reported to Access.The record seems to disappear, but when rereading the table (or recordset), the 'deleted' record reappears. Yes, this is a known problem with the current ODBC driver. There are patches that fix this. Who's in charge of the ODBC driver? Patches are fine, but I'm sure there are many of us that cannot compile under windows, myself included. When can we get RI violation reporting included in the 'real' ODBC driver? I don't like to rely on a special 'hacked' version of something in a production environment. Thanks -Cedar
hi, there! On Tue, 10 Oct 2000, Cedar Cox wrote: > > The bad news, whenever I delete or update a record causing error in referential integrity, no error is reported to Access.The record seems to disappear, but when rereading the table (or recordset), the 'deleted' record reappears. > > Yes, this is a known problem with the current ODBC driver. There are > patches that fix this. > > Who's in charge of the ODBC driver? Patches are fine, but I'm sure there > are many of us that cannot compile under windows, myself included. When > can we get RI violation reporting included in the 'real' ODBC driver? I > don't like to rely on a special 'hacked' version of something in a > production environment. by the way -- backend protocol does not return numeric error codes so ODBC driver cannot set SQLCODE's. (SQLCODE returned by ODBC driver is always empty). So applications that use ODBC can not distinguish for example, referential integrity errors from backend fatal errors or errors in SQL statement. Of course, it is possible to examine string error messages which are returned by backend by this is rather hackish and there is no evidence that they will not change in future. /fjoe